Dual pathway for metabolic engineering of Escherichia coli to produce the highly valuable hydroxytyrosol

One of the most abundant phenolic compounds traced in olive tissues is hydroxytyrosol (HT), a molecule that has been attributed with a pile of beneficial effects, well documented by many epidemiological studies and thus adding value to products containing it. Strong antioxidant capacity and protection from cancer are only some of its exceptional features making it ideal as a potential supplement or preservative to be employed in the nutraceutical, agrochemical, cosmeceutical, and food industry. The HT biosynthetic pathway in plants (e.g. olive fruit tissues) is not well apprehended yet. In this contribution we employed a metabolic engineering strategy by constructing a dual pathway introduced in Escherichia coli and proofing its significant functionality leading it to produce HT. Our primary target was to investigate whether such a metabolic engineering approach could benefit the metabolic flow of tyrosine introduced to the conceived dual pathway, leading to the maximalization of the HT productivity. Various gene combinations derived from plants or bacteria were used to form a newly inspired, artificial biosynthetic dual pathway managing to redirect the carbon flow towards the production of HT directly from glucose. Various biosynthetic bottlenecks faced due to feaB gene function, resolved through the overexpression of a functional aldehyde reductase. Currently, we have achieved equimolar concentration of HT to tyrosine as precursor when overproduced straight from glucose, reaching the level of 1.76 mM (270.8 mg/L) analyzed by LC-HRMS. This work realizes the existing bottlenecks of the metabolic engineering process that was dependent on the utilized host strain, growth medium as well as to other factors studied in this work.


Introduction
Hydroxytyrosol (3,4-Dihydroxyphenylethanol; HT) is a mono-phenolic compound traced in olive fruits [1] and tissues [2], in extracted olive oil [3], or even in olive mills waste waters [4]. It shows a broad spectrum of biological properties due to its strong antioxidant and radical- tyrosine as a precursor, utilizing the following genes: an aromatic acetaldehyde synthase (AAS), an aldehyde reductase (ALR) and a tyrosinase (TYR) ( Table 1). Various efforts were undertaken to optimize this heterologous HT production. We marked each factor's specific effect on the metabolic flux and on the final HT yield. To achieve this, we engineered regular strains [BL21(DE3) and HMS174(DE3)], expressing genes not only to produce HT, but also for the endogenous tyrosine overproduction, thus leading to highest titers of HT directly from glucose. Such a metabolic effect, permitted the redirection of primary metabolic flux from externally added glucose to a tyrosine pool that proved to be saturating the first enzymatic step (Fig 1) of the engineered pathway from tyrosine to HT. Moreover, the addition of a selected ALR (ALR-K, Table 1), overexpressed along with the other abovementioned genes, (Fig 1B), led to additional increase in HT production efficiency.

Materials and DNA extraction
All chemicals used were purchased from Sigma-Aldrich unless otherwise stated. Genomic DNA extractions were accomplished from overnight cultures with Qiagen Blood & Tissue kit according to manufacturer instructions. PCR amplified fragments and digested fragments Due to the dual specificity of the utilized aromatic acetaldehyde synthase (AAS) cloned from parsley, and to tyrosinase (TYR), cloned from Ralstonia solanacearum, a dual pathway is generated. The "monophenol route" (blue) pathway portrays the enzymatic route for HT biosynthesis deriving from the decarboxylation-deamination of tyrosine, while the "diphenol route" (red) pathway portrays the HT route deriving from the hydroxylation of tyrosine to L-3,4-dihydroxyphenylalanine (DOPA). Gene-respective enzyme abbreviations: AAS, aromatic acetaldehyde synthase. ALR, aldehyde reductase. AroA, 5 Table 1). Bold abbreviations denote the genes that were utilized in this study.
were purified with Macherey-Nagel NucleoSpin Gel and PCR Clean-up kit. Restriction enzymes were purchased from Minotech, Crete, Greece or New England Biolabs. Taq DNA polymerase was provided from Kapa Biotech while the proof-reading polymerase used for all the cloning processes (Phusion) was obtained from New England Biolabs. For SDS PAGE the PiNK prestained protein ladder was used (Nippon Genetics).

Genetic material, microbial hosts and cloning vectors
All gene descriptions are listed in Table 1. The TYR gene was cloned from the strain GMI1000 of Ralstonia solanacearum [32], the AAS gene was cloned from root cDNA of Petroselinum crispum (parsley). The yahK and yqhD genes that encode for aldehyde reductases were cloned from E. coli genomic DNA [30]. For further convenience, we will refer to these two genes with the abbreviations ALR-K and ALR-D respectively (Table 1). All genes involved in the tyrosine overproduction pathway (Fig 1A) were kindly donated by Prof. J. D. Keasling [29].  [32] All cloning and sub-cloning steps were done using the Novagen pRSF and pET dual expression (Duet) vectors, all comprising the pET protein expression system. It was developed for the cloning and expression of recombinant proteins in E. coli. Target genes were cloned into Table 2. Metabolically engineered bacterial plasmids (see Table 1  The asterisks in aroG � and tyrA � refer to the feedback-resistant variants of aroG and tyrA, respectively [29]. OP designation stands either for a tyrosine (Fig 1A) or a hydroxytyrosol (HT, Fig 1B)  pET plasmids under the control of a strong bacteriophage T7 transcription and (optionally) translation signals. The Escherichia coli strain used for accomplishing the cloning strategy was the DH10B while the host bacterial strains used for expression experiments were the BL21 and HMS174 (Table 2). BL21 and HMS174 were evaluated for use as hosts because the former belongs to the B line of E. coli, while the second belongs to the K12 E. coli line [34].

Software tools
The prediction of twin-arginine translocation signal peptide in R. solanacearum tyrosinase was assessed with the PRED-TAT web integrated tool (ww.compgen.org/tools/PRED-TAT/submit) utilizing Hidden Markov Models [35]. Primer design and in-silico DNA manipulations were completed with SnapGene (http://www.snapgene.com/).

Engineering of expression vectors with genes involved in hydroxytyrosol biosynthesis
Construction of tyrosinase and aromatic acetaldehyde synthase expression vectors. All engineered plasmids constructed in this study were based on Duet expression vectors. The construction strategy and steps for each engineered plasmid used in this study are depicted in Fig 2. Genes were either cloned by PCR utilizing appropriately designed oligonucleotides from various sources (S1 Table). To confirm each gene's identity the PCR cloned fragments were sequenced and compared in silico with the NCBI deposited sequences.
All plasmid constructions used in this study are shown in Table 2 and Fig 2. The TYR gene was amplified from R. solanacearum genomic DNA with the primer pair RsTYRfw2/RsTYRrv2 (S1 Table). The amplified DNA fragment was digested with the restriction enzymes NdeI and XhoI that flank the fragment and was inserted into the respective sites of an empty pRSF Duet vector (Novagen). The resulted vector was named pRSF-RsTYR. The AAS gene from P. crispum (parsley) was amplified from parsley root cDNA, where pathogen responsive genes [36] are usually active [37], with the primer pair PcTDC2bfw/PcTDC2bfw2 (S1 Table). The amplified DNA fragment was digested with the restriction enzymes NcoI and BamHI and inserted into the respective site of the pRSF Duet vector generating the pRSF-PcAAS expression vector. To create the double gene expressing pRSF vector, the pRSF-RsTYR plasmid was digested with HindIII/ XhoI enzymes to excise the RsTYR expression module and inserted into the similarly digested pRSF-PcAAS, thus creating the pRSF-PcAAS-RsTYR dual expression vector.
Construction of aldehyde reductase expression vectors. Two aldehyde reductase (ALR) genes from E. coli BL21(DE3) (ALR-D, and ALR-K) were cloned into pRSF Duet and pCDF Duet vectors. The ALR-D and ALR-K were amplified with the yqhD-fw1/yqhD-rv1 and yahK-fw1/yahK-rv1 primer pairs respectively (S1 Table) and were cloned into pCDF Duet vectors after digestion of both PCR and vector with NcoI/PstI. The resulted plasmids were named pCDF-ALR-D or pCDF-ALR-K respectively (S1 Table).

In-vivo expression and growth optimization experiments
All genes referred in Table 1 were used primarily in in-vivo experiments, to check their optimal expression as well as their function. The latter was achieved by measuring their activity of their respective enzyme product activity through the quantity of their equivalent end-product biosynthesis.
In feeding experiments the LB-M9 protocol was used according to which a 5 mL starter culture was set in LB (1% tryptone, 0.5% yeast extract, 1% NaCl) from freshly transformed E. coli with appropriate plasmids and left to grow overnight at 37˚C. The next day, 50 mL of LB were inoculated with the overnight culture to OD 600 = 0.05-0.1 and were left to grow to an OD 600 = 0.4-0.6. At that point isopropyl β-D-1-thiogalactopyranoside (IPTG) was added at 1 mM concentration to induce the expression of the heterologous gene(s). After an induction time of 3 hours the cells were harvested by centrifugation (6000 g, 10 min), suspended in M9 medium (22.3 mM Na 2 HPO4�7H 2 O, 22 mM KH 2 PO 4 , 8.6 mM NaCl, 18.7 mM NH 4 Cl, 1 mM MgSO 4 , 25 μM CuSO 4 , 0.1 mM CaCl 2 , 10 nM thiamine, 55 mM glucose) and appropriate substrate was added. 50 mL fermentations were performed in 250 mL Erlenmeyer flasks with orbital shaking at 200 rpm and 30˚C. For the analysis of the substrate consumption and the product biosynthesis, 1 mL samples were collected and kept at -20˚C until their electro-chromatographic analysis.
For HT production directly from glucose the simple M9 protocol was followed according to which, similarly to LB-M9, a 5 mL starter culture was set in LB with appropriate plasmids and left to grow overnight at 37˚C. The next day, 50 mL of LB were inoculated with the overnight culture to OD 600 = 0.05-0.1 and were left to grow to an OD 600 = 0.4-0.6. The cells were harvested by centrifugation (6000 g, 10 min), washed with M9 broth containing the basic salts and resuspended in M9 broth as in LB-M9 protocol. For the protein induction, IPTG was added at 50μM concentration. 50 mL fermentations were performed in 250 mL Erlenmeyer flasks with orbital shaking at 200 rpm and 30˚C. For the analysis, 1 mL samples were collected and kept at -20˚C until their electro-chromatographic analysis.

Tyrosinase and aromatic acetaldehyde synthase in-vitro enzymatic determination
To obtain bacterial cellular extracts enriched in the expressed proteins, freshly transformed BL21 clones with the appropriate expression vector were grown at 37˚C overnight in 5 mL LB. The next day 50 mL of LB in a 250 mL Erlenmeyer flask were inoculated with 1 mL of the overnight culture and left to grow for 3 hours to reach and OD 600 of about 0.5. At that point the inducer IPTG was added to a final concentration of 1 mM. Cells were left to grow for 20 hours and then the cells were harvested by centrifugation (3000 g, 10 min, 4˚C) and washed with 50 mL of cold 20 mM Tris�HCl pH 7, harvested again and resuspended in 5 mL of the same buffer. The cells were disrupted by sonication with a Braun Labsonic U sonicator (5-min treatment at a relative output power of 0.5 with 0.5 duty period). The homogenate was centrifuged at 20000g for 30 min, and the supernatants were used for enzymatic activity determinations.
TYR activity was determined as previously described [38] for tyrosine hydroxylase. Briefly, TYR activity was carried out at room temperature in an assay (1 ml) containing 0.05% SDS, 20 mM phosphoric buffer pH 5, 1 mM tyrosine, 25 μM CuSO 4 . The reaction was initiated upon the addition of a crude extract containing 500 μg of total proteins (described above) estimated with the Bradford assay [39]. Successive samplings of 100 μl at 0, 10, 20, 30, and 60 minutes after reaction start were analyzed in CE.

Electro-chromatographic analysis
Qualitative and quantitative analysis of samples were carried out in an Agilent Capillary Electrophoresis (CE) G1605 system coupled with a Diode Array Detector (DAD). A bare fused silica capillary column was used with effective length of 50 cm and inner diameter of 75 μm. The analysis was carried out in several steps: preconditioning, inlet and outlet buffer replenishment, buffer customization, injection and analysis. The preconditioning step involved an initial flush with 1 N sodium hydroxide for 2 min followed by a flush with analysis buffer (25 mM sodium tetraborate decahydrate, pH 9.2) for 4 min. Buffer customization included the application of 20 kV voltage for 2 min before each analysis The injection step comprised a 50 mbar pressure for 2 sec and the analysis was performed at 29 kV for 10 min, time sufficient for maximum resolution of the compounds analyzed in this study. The temperature of the column cassette was kept constantly at 25˚C. Replenishment and buffer customization steps were necessary to eliminate poor reproducibility of the peaks resulted from electrolytic phenomena in the running buffer and compound instability. Compounds under study were identified by matching the retention time, UV-absorbance spectrum, and co-chromatography with authentic chemicals. Calibration curves were obtained with authentic compound solutions of various concentrations.
All experiments were performed at least 3 times unless otherwise stated. Data of each experiment were used to calculate the mean and the respective standard deviations.

Mass spectrometry analysis of HT production
Further qualitative and quantitative analysis was performed on the most efficient metabolically engineered E. coli strain that produced the highest concentration of HT. The E. coli strain HMS174-TYR OP -HT OP -ALR-K (Table 2) was grown in 5 L flask filled with 1 L of growth medium. After 48 h of growth, samples of metabolically engineered E. coli strain were analyzed through mass spectrometry.
For the qualitative and quantitative monitoring of the E. coli strains an Accela HPLC system (Thermo), coupled to a hybrid LTQ Orbitrap Discovery XL mass spectrometer (LC-HRMS) and equipped with an electrospray ionization source (ESI) was used. The separation was conducted using a gradient elution system consisted of water with 0.1% (v/v) formic acid (A) and acetonitrile (B). The elution started with 5% B and reached 50% in 5 min. After 2 min the system returned to the initial conditions and stayed for 5 min for equilibrium of the column. The flow rate was set at 0.4 ml/min and the total run time was 15 min. For the chromatographic separation a Fortis C-18 (1.7 μm, 150 x 2.1 mm) column was used heated at 40˚C. The injection volume was 10 μl and samples were maintained at 10˚C during analysis. Ionization was achieved in negative ion mode (ESI-) at 350˚C. The mass spectrometric parameters were: sheath gas and aux gas flow rate 40 and 10 units respectively; capillary voltage -40 V and tube lens -69 V. The mass range was adjusted from 113 to 1000 m/z.
In order to quantify HT, calibration curves were built using six different concentration levels of the analyte. The selected levels were 1 μg/ml, 2 μg/ml, 4 μg/ml, 6 μg/ml, 8 μg/mL and 10 μg/ml. 2.4-Dinitrophenol was used as internal standard (IS) in the concentration of 0.3 μg/ml. The construction of the calibration curve based on the ratio area of HT/IS versus the concentration of HT. Linearity was evaluated by coefficient of determination (R 2 ) from the linear regression using the least squares equation. All concentration levels were measured in triplicates. Blank samples were injected every three injections in order to avoid carryover effect of the different concertation levels. Data were acquired, analyzed and processed with the Thermo Xcalibur 2.1 software.
For the bacterial strain extraction, 225 mL of cell culture were centrifuged at 3500 rpm for 20 min. The supernatant was extracted with the adsorption resin Amberlite XAD7 after overnight treatment, in order to obtain an extract being enriched in phenolics. Methanol was used as extraction solvent. The obtained methanol extract was dried under vacuum until dry and reconstituted to 100 μg/mL for the HPLC-ESI-HRMS analysis.

Expression of genes involved in the biosynthesis of hydroxytyrosol
Ralstonia solanacearum tyrosinase. The hydroxylation of the phenolic ring of tyrosine or tyrosol to form DOPA or HT respectively may be catalyzed by the action of the R. solanacearum TYR (Table 1). BL21 cells bearing the RsTYR (BL21-pRSF-RsTYR, Table 2), in induced conditions were initially checked for their ability to induce the expression of the TYR protein (S3A Fig). As negative controls, BL21 cells were used transformed either with empty pRSF vector (in non-induced conditions) or with pRSF-RsTYR (also in non-induced conditions). It was obvious that an intense band (see arrow in S3A Fig), that did not exist in negative controls, was present above 43 KDa, corresponding to the TYR protein size of about 50 kDa [32]. Crude extracts from these cultures (non-induced BL21-pRSF, non-induced BL21-pRSF-RsTYR, and induced BL21-pRSF-TYR) were tested for TYR activity in a colorimetric assay since the dual activity of the enzyme oxidizes DOPA to the orange dopachrome and eventually to blackish melanins ( Fig  3B). Conclusively, both the extracts from the induced and non-induced conditions were able to transform tyrosine to DOPA (and eventually to melanins) indicating from one hand the activity of the enzyme and on the other hand the leaky expression of TYR in non-induced conditions. Subsequently, TYR activity was verified by in-vitro ( Fig 3A) as well as colorimetric assays (Fig 3B). A time course experiment following tyrosine and DOPA certified that tyrosine was promptly converted to DOPA (Fig 3A). In all cases the color of the reaction was getting black indicating the melanization of DOPA through the diphenolase activity of TYR.
Furthermore, in-vivo assays for the conversion of 2 mM tyrosine or tyrosol to DOPA or HT respectively were set up (Fig 4). In induced conditions, the substrate (tyrosine or tyrosol) was converted to DOPA or HT respectively 2 h after the addition of the substrate (Fig 4B and 4D). When the cultures were kept under non-induced conditions, no consumption of the substrate was recorded, or the consumption was minimal. A slight increase of the products was also recorded (Fig 4A and 4C).
According to PRED-TAT algorithm [35] and the integrated web tools (http://www. compgen.org/tools/PRED-TAT), the N-terminal of R. solanacearum TYR (497 aa) was found to bear a 31 aa peptide for translocation of the protein to the periplasm [40] (S1 Fig) with reliability score of 0.998 (new model, max value 1).
Petroselinum crispum aromatic acetaldehyde synthase. Beyond the hydroxylation of the phenolic ring, another essential reaction is the decarboxylation-amination of the aromatic amino acid by the respective AAS. Newly transformed BL21 cells bearing the PcAAS  Table 2) in induced conditions were checked for their ability to induce the expression of the AAS (S3B Fig). As negative control, BL21 cells transformed with pRSF-PcAAS, in non-induced conditions, were used. As with TYR producing BL21 cells, a band with much higher intensity was present at about 50 KDa [36], corresponding to the AAS protein.
In-vivo assays for the evaluation of AAS were set up. Induced cultures were supplemented either with tyrosine or DOPA and time course samples were analyzed in CE (Fig 5). Either substrate, tyrosine or DOPA, was converted to tyrosol or HT respectively, with the latter substrate (DOPA) to be consumed at about 5 hours after substrate addition. Similarly to TYR invivo assays, the cultures under non-induced conditions presented minimal consumption of the substrate (Fig 5A and 5C) while the induced conditions exerted the maximal conversion efficiency (Fig 5B and 5D).

HT production from engineered E. coli strains and IPTG effect
The proof that RsTYR and PcAAS genes were active, was followed by the combined in-vivo expression. Cultures of BL21 cells transformed with pRSF-PcAAS-RsTYR (Table 2) were set up to evaluate their ability to convert supplemented tyrosine into HT in both induced and non-induced conditions (Fig 6A). As a rule, 1 mM of tyrosine was supplemented. The strain was evaluated on different IPTG concentrations; 0, 50, 100, 250, 500, 750, and 1000 μM. A first note was the substrate loss to melanin accumulation through the oxidation of the tyrosine derived DOPA through dopachrome observed by the blackening of the media. As the IPTG   concentration decreased from 1 mM to 0 mM, differences were observed in the produced analytes; the intermediate tyrosol was lowering while the concentration of HT increased. Furthermore, in the lower IPTG concentrations (50 and 100 μM) DOPA was not traced. When no tyrosine was added to the system a minimal amount of HT (0,21 mM) was produced in induced conditions or no HT was produced in non-induced conditions. The consumption of tyrosine in non-induced conditions (0 mM IPTG, Fig 6) further propose the leaky activity of TYR and AAS. The negative control BL21 cells that did not carry any of the HT pathway genes, did not consume the exogenous supplied tyrosine. Consequently, 50 μM of IPTG was regarded as the optimal concentration of the inducer and was used throughout our experimental conditions unless stated otherwise.

Hydroxytyrosol production from engineered tyrosine-overproducer E. coli strains
Instead of using an E. coli system that is necessary to be supplemented with the precursor amino acids tyrosine or DOPA to produce HT, a biological system designed to overproduce tyrosine was further engineered by introducing the pS4 and pY3 plasmids to BL21 and HMS174 hosts, for the production of HT directly from glucose (BL21-TYR OP -HT OP and HMS174-TYR OP -HT OP strains, Table 2). Experiments with these strains showed non-consistent tyrosine production, while HT levels remained similar reaching the level of 0.21 mM in HMS174 strain (Fig 7). This was much an unexpected result, since there was a noticeable difference on tyrosine pool, between the two strains. While in BL21-TYR OP -HT OP appeared that all internally overproduced tyrosine was at nearly non-existing levels indicating that it was all consumed in primary metabolism (Fig 7A), in HMS174-TYR OP -HT OP , tyrosine levels reached 1.3 mM. Similar biosynthetic levels were obtained for the intermediate metabolites DOPA and tyrosol in HMS174-TYR OP -HT OP , however, leading to still low HT level. This metabolic event indicated that a probable biosynthetic bottleneck existed in the conversion of the internally produced tyrosine to HT.
Auxiliary aldehyde reductase genes. The biosynthetic pathway of this work utilizes the AAS from P. crispum ( Table 1) that possess dual specificity in respect to the amino acid that is accepted as substrate. It can convert tyrosine or DOPA to 4-hydroxyphenylacetaldehyde (HPAA, monophenol route) or 3,4-dihydroxyphenylacetaldehyde (DHPAA, diphenol route) respectively ( Fig 1B). The next step should be the reduction of the phenylacetaldehydes to the respective alcohols, tyrosol or HT. E. coli is an organism that possesses several ALRs that may drive the abovementioned conversions [27]. We therefore checked whether the presence of an extra overexpressed ALR could confer any advantage to the engineered system. The ALRs evaluated were the ALR-K and the ALR-D [30] both of which are present in the genomes of BL21 and HMS174 strains. The over-expression of ALR-D along with the HT biosynthesis genes did not give any advantage to the production system, either expressed in BL21 or HMS174 strain. Though, this was not the case for the over-expression of ALR-K. It gave a decrease of 63% in the production of HT when expressed in BL21-TYR OP -HT OP (Fig 8A) while it gave an increase of 386% when expressed in HMS174-TYR OP -HT OP (Fig 8B) reaching a concentration of 1.02 mM (157.2 mg/L) when analyzed in CE-DAD system. All the follow up experiments continued with optimization of this highest HT titer producing strain.
Optimal analysis of the highest HT producer strain through Mass Spec analysis. The strain HMS174-TYR OP -HT OP -ALR-K (Table 2), carrying the HT overproducing pathway (involving the genes RsTYR, PcAAS and EcALR-K, Table 1), as well as all the genetic accessories to overproduce tyrosine ( [29], Table 1), was further analyzed by an LC-MS-MS type mass spectrometry apparatus. In this strain, the precursor of HT pathway, tyrosine (Fig 1) was internally produced by feeding the strain with glucose. This strain grown under the M9 protocol, showed the highest HT production when samples were analyzed from 50 mL cultures ( Fig  8B) with a CE-DAD system. However, it was important to find whether there was any effect on HT production when grown on large volume medium (1 L) and analyzed by mass spectroscopy. Indeed, the MS analysis showed HT production at much higher concentration level.
After the purification and enrichment process of the strain HMS174-TYR OP -HT OP -ALR-K using adsorption resins, the dried weigh of the methanol extract was estimated at 4.2 mg/mL cell culture. In order to determine the HT levels, HPLC-ESI(-)-HRMS was employed. Fig 9 illustrates the obtained base-peak chromatogram of the extract after resin treatment together with the corresponding mass spectrum of HT. As it is highlighted in red, HT was eluted at 2.94 min and the dinitrophenol internal standard at 6.71 min. Other peaks were also detected, mainly representing intermediate compounds of the metabolic pathway. At the mass spectrum (Fig 9, smaller graph) the ions corresponding to HT were presented. The high accuracy and resolving power of the Orbitrap analyzer as well as the use of reference standard enabled the unambiguous identification of HT in the complex mixture [41]. Together with the pseudomolecular ion of HT ([M-H] -), other ions corresponding to its adducts with formic acid ([M+-FA-H] -) and its dimer ([2M-H] -) were also detected. Using Thermo Xcalibur 2.1 software and extraction ion method (XIC), the quantification of HT took place using the linear regression method. A standard calibration curve was built described by the equation y = 0.1674x + 0.1929 while coefficient of determination was calculated (R 2 = 0.9932), indicating the good linearity achieved. Based on the equation, the level of HT was calculated at 270.8 mg/L of cell culture.

Discussion
In this project we succeeded to modularly engineer both the primary and secondary metabolism of E. coli to produce HT. We chose a rational approach for its production [16] where the primary metabolism was engineered in order to boost tyrosine production that is subsequently utilized for the biosynthesis of HT through the engineered secondary metabolic machinery.
The HT molecule bears a benzene (phenolic) ring linked with an alcoholic chain with two carbon atoms (Fig 1). Since all phenolic compounds encountered in organisms emanate from the aromatic amino acids via the shikimic acid pathway [42,43], we assumed that this is also the case for HT biosynthesis. From the three aromatic amino acids of the primary E. coli metabolism (phenylalanine, tyrosine, tryptophan), HT resembles most with tyrosine (Fig 1) and could be biosynthesized by a hydroxylation, a decarboxylation, a deamination, and finally a reduction (Fig 1B).
Thus, we chose to utilize the TYR gene from R. solanacearum (Table 1) for the hydroxylation of tyrosine (or tyrosol) [32] and the AAS gene from parsley [36]. The TYR enzyme beyond the ortho-hydroxylation of the phenolic ring also oxidizes the hydroxyl groups to quinones [24,32]. On the other hand, the AAS enzyme has also the advantage that beyond the basal decarboxylation reaction, catalyzes an extra deamination step that leads to the production of the corresponding phenylacetaldehyde [31]. The activity of R. solanacearum TYR was confirmed in in-vitro (Fig 3) as well as in in-vivo experiments on tyrosine (Fig 4A and 4B) and on tyrosol (Fig 4C and 4D) while the activity of the parsley AAS was confirmed in in-vivo experiments where reactions were supplemented with either tyrosine (Fig 5A and 5B) or DOPA (Fig 5C and 5D). The reduction of the produced aromatic phenylacetaldehyde (4-hydroxyphenylacetaldehyde produced from the monophenol route or DHPAA produced from the diphenol route) was accomplished by internal ALRs that both BL21 and HMS174 E. coli strains possess [27]. The simultaneous expression of RsTYR and PcAAS was evaluated by the ability of the system to transform the supplemented tyrosine to HT (Fig 6) and proved to be a successful strategy for the robust manufacturing of the high-valuable compound HT.
The dual specificity of both utilized enzymes, TYR acting on tyrosine or tyrosol and AAS acting on tyrosine or DOPA, unwraps two possible metabolic routes for the production of HT; one through the decarboxylation-deamination of DOPA (diphenol route, Fig 1B) for the production of DHPAA, and one through the decarboxylation-deamination of tyrosine for the production of 4-hydroxyphenylacetaldehyde (monophenol route). Both parallel occurring metabolic routes, lead to the production of HT.
The analysis of RsTYR in in-vitro and in-vivo activity experiments (Fig 3 and Fig 4), where the precursor molecules were exogenously supplemented, revealed that the enzyme was active in both induced and uninduced conditions. More interestingly, it performed quickly, within the first 20 min with the in-vitro crude E. coli extract. This was advocated by the identification of DOPA in early time points upon the supplementation of tyrosine. The in-vivo rapid conversion of the medium supplemented tyrosine could be explained by the finding that RsTYR was transferred to the medium through the periplasmic space of the host as it possesses a TAT signal peptide (S1 Fig). The extracellularly produced DOPA was not directly available to AAS, that was intracellularly expressed. This differential compartmentalization may explain the reason why extracellular DOPA could not be easily converted to HT and was wasted as melanized byproducts due to exposure of supplemented tyrosine to RsTYR, leading to the oxidation of DOPA accompanied by the blackening of the medium (Fig 3B). Consequently, the late increase in HT is assumed to be produced through the newly synthesized endogenous tyrosine and the monophenol route (Fig 1). In that case, the intracellularly AAS-synthesized tyrosol was transferred outside of the cell before its transformation to HT by TYR.
The ability of the engineered E. coli to bio-transform the supplemented precursor tyrosine into the valuable HT as well as the fact that tyrosine was the limiting factor for producing HT (since its production rate was rapidly reduced to zero after tyrosine consumption; Fig 5B), urged us to evaluate the case of producing HT from a tyrosine over-producing strain. The HMS174 strain proved more efficient to produce tyrosine than BL21 as the former was able to convert the supplemented glucose to 6.23 mM of tyrosine. The efficient biosynthesis of tyrosine was necessary to avoid the exogenous supplementation of tyrosine and thus making the biological production of HT more economically attractive. To this extend, E. coli strains containing the modules for a) the biosynthesis of shikimic acid (pS4), b) the production of tyrosine from shikimic acid (pY3), and c) the production of HT from tyrosine, were utilized (BL21-TYR OP -HT OP and HMS174-TYR OP -HT OP ) to produce HT. As was earlier mentioned, the supplementation of tyrosine to BL21-RsTYR-PcAAS or HMS174-RsTYR-PcAAS strains led to intermediate product formation due to spatial separation of the TYR and AAS activities. The channeling of intracellularly produced tyrosine to the monophenol route of HT pathway (Fig  1) resulted in the formation of tyrosol that was eventually excreted into the medium where it was converted to HT. However, both engineered strains resulted to non-equimolar levels of HT, suggesting a metabolic burden although the HMS174 proved a much higher tyrosine producer than BL21-HT OP .
Here by, we tried to translate the evidence to find the preferable route toward the biosynthesis of HT (monophenol or diphenol, Fig 1). When we studied each enzyme activity alone (either TYR in Fig 4 or AAS in Fig 5), there was in-vivo evidence that TYR was able to consume its substrates faster (tyrosine or tyrosol, within less than 2 h as compared to AAS consuming tyrosine or DOPA that lasted at least 5 h) as well as tyrosine was converted to DOPA by AAS but not at equimolar ratio (Fig 5). The abovementioned evidence suggests a delay and inefficiency in AAS activity, either alone or along TYR, that results in tyrosol and subsequently to HT accumulation (Fig 6), even at the optimal IPTG concentration. This happened most probably because the active center of TYR was either saturated with tyrosine or with DOPA and/or HT produced from the red path, causing the active TYR to exert diphenolase activity that may lead to partial inactivation due to the quinones produced [24,44]. It should also be noted that under conditions that AAS and TYR were absent, there was no significant tyrosine consumption showing that its metabolism when provided in excess was due to the heterologous pathway. Thus, we conclude that when the system is not supplemented with any precursor molecule the metabolic flow starts from glucose and ends up to HT through the diphenol route ( Fig 1B).
The intermediate tyrosine, DOPA and tyrosol build-up in HMS174 strain, in contrast to BL21 (Fig 7A and 7B), shows its potential as an HT over-producer that is kept low due to the presence of a bottleneck in the pathway. This further showed that the initial hypothesis for the advantage of the dual function of our proposed pathway was not quite exact as the native E. coli ALR appeared not to be as active as expected even though we attempted to optimize growth and induction conditions. Indeed, this limiting step was overcome by the overexpression of an alternative ALR gene that was cloned from E. coli and led to substantial increase to the production of HT. Indeed, the question as to whether the activity of a native ALR gene could be the limiting factor in the HT production, was answered by the over-expression of two E. coli ALR genes, the ALR-D and the ALR-K along with the HT biosynthetic genes (Fig 1B) and the two modules (Fig 1A, [29]) responsible for the tyrosine overproduction. These two ALR genes were selected based on earlier evaluation of their ability to convert similar aldehydes into alcohols [30,45]. Interestingly, while the over-expression of ALR-D did not cause any further improvement of HT production, the over-expression of ALR-K increased the production of HT by 386% reaching a concentration of 1.02 mM (157.2 mg/L, Fig 8B). When the ALR-K was expressed in the respective BL21 strain, a reduction of 63% was estimated possibly because toxic intermediates were build-up. This means that the involvement of the overexpressed ALR-K resolves the biosynthetic bottleneck problem resulting in equimolar production levels of HT as compared to tyrosine initial levels ( Fig 8B). Moreover, when ALR-K was functioning, the levels of DOPA and tyrosol remained minimal showing that the biosynthetic flow from tyrosine to HT was running properly; the endogenously produced tyrosine was subjected to the activity of the AAS (and the activity of the ALR) for the production of tyrosol and eventually to its transformation into HT through the hydroxylation of tyrosol by the TYR (monophenol route, Fig 1).
To slightly expand our thinking, we furthermore assumed that some intermediate products were consumed to alternative routes not for the production of HT. One of these routes is most likely to involve the function of feaB, producing the respective acid derivatives and thus removing precursors and substrates that would otherwise lead to HT production. The negative effect of feaB to remove intermediate metabolites away of the pathway of interest has been shown before [27,28]. Such acid production may affect the activity of AAS and TYR, as showed by the decrease and/or accumulation of their respective products DOPA and/or tyrosol. Suggested deletion of feaB gene previously resulted in further increase of HT production [27].
The expression of the ALR-K is relieving the pathway from the effect of a possibly deleterious accumulating product. Its function is to convert at higher rates hydroxyphenyl-acetaldehyles produced to hydroxyphenyl-alcohols, such as tyrosol (monophenol route, Fig 1A) or hydroxytyrosol (diphenol route, Fig 1B). Given the fact that both parental strains (BL21 and HMS174) possessed the phenylacetaldehyde dehydrogenase gene (feaB) thus acting as hydroxyphenylacetic acid producers (Fig 1), led us to hypothesize that feaB must have a faster turn over than the natively existing ALR, which favored the conversion of HPAA and DHPAA to the respective acetic acid derivatives [30].
Further LC-HRMS analysis on scaled-up culture volume (1 L) of HMS174-TYR OP -HT OP -ALR-K strain detected higher HT production reaching 1.76 mM (270.8 mg/L, Fig 9). This increase in HT production may also be due to the stationary phase protein overproduction that is a fundamental capability of E. coli [46], which has been shown to lead to protein expression with higher protein solubility and thus higher enzyme activities [47].
In contrast to previous attempts to produce HT directly from glucose [27,28], we utilized different gene sources encompassing different activities and different host strains to show the effect of the coexistence of di-and mono-phenol route. The final titer of this metabolic construction was about 22 times higher than that of Satoh et al. [27] and 1.3 times higher than that of Chung et al. [28] (Table 3). The molarity of produced HT was 3.5 times less than the herein reported tyrosine production (6.23 mM) or 6.8 times less than the tyrosine reported to be produced by the tyrosine overproducer strain [29], a fact that dictates space for further optimization.
At the time of submission, two new articles drawn our attention for the biosynthesis of hydroxytyrosol in E. coli, authored by Li et al. [51] and Choo et al. [50]. The performance of their system is of interest due to the efficacy they achieved, especially for the case of Li et al. [51] who achieved a titer of 647 mg/L (4.2 mM, Table 3). They ended up in such a high concentration following a different approach, directing the metabolic flow from 4-hydroxyphenylpyruvate to 4-hydroxyphenylacetaldehyde, tyrosol and finally to HT through the action of a ketoacid decarboxylase, an alcohol dehydrogenase and a 4-hydroxyphenylacetic acid 3-hydroxylase, respectively. The pathway was overloaded with 4-hydroxyphenylpyruvate by the action of an aromatic-amino-acid aminotransferase. Their pathway proved more efficient in terms of HT production, utilizing four enzymes in contrast to our strategy that utilized fourteen; this may be an explanation of why they succeeded higher efficacy. Metabolic engineering of Escherichia coli to produce hydroxytyrosol Concluding, the genetically tractable microbe E. coli provides a supreme platform for the combinatorial biosynthesis of plant natural products utilizing heterologous genes from various sources. Microbial production of plant natural products is a promising alternative to traditional methods. Here, we presented the numerous efforts made for the optimization of HT production directly from glucose utilizing a dual-pathway approach that drove the system to a HT concentration of 270.8 mg/L within 48 h. Though, further optimization efforts are ongoing to increase the final titers particularly, important for the entry of the system to the industrial production, such as the use of an AAS variant that present monospecific substrate activity that will be coupled with a TYR variant with less [44] or absent diphenolase activity or the chromosomal integration of the implicated genes. The latter may lead to the removal of antibiotics from the production plan that will further help to increase the HT purification efficiency.
Supporting information S1 Table. List of primers used for gene clonings included in this study. In the first lane total proteins from E. coli BL21-pRSF in noninduced conditions were loaded. In the second, total proteins from BL21-pRSF-RsTYR in non-induced conditions were loaded while in the third lane total proteins from BL21-pRSF-RsTYR in induced conditions were loaded. In the last lane the PiNK prestained protein ladder was loaded. The arrow in the protein marker helps to estimate the size of the protein band. Lower part, colorimetric assay with the protein crude extracts to assess the activity of tyrosinase as described in Material and Methods. (B) Protein expression in E. coli to certify the AAS expression. In the first lane the PiNK prestained protein ladder was loaded. In the second and the third lanes total proteins from BL21-pRSF-PcAAS in induced and non-induced conditions were loaded respectively. The arrows in the protein marker help to estimate the size of the expressed protein band. (TIF)